Abstract

Multi-vendor report of running SNA protocols over TCP/IP using the Data Link switching approach. This test included the following solutions: 3Com NETBuilder II, and Cisco Systems 7010 Multiprotocol Router.

Testing involved running SNA APPC traffic from Token Ring LAN to Token Ring LAN via a WAN link.

This Tolly Group executive summary examines early “pre-standard” Data Link Switching (DLSw) implementations designed to solve a longstanding problem in token-ring internetworking: preserving IBM SNA/LLC2 session integrity across congested WAN links. Traditionally, remote PCs accessed mainframe resources across token-ring bridges using source-route bridging, but that approach offered no protection for LLC2 sessions when WAN congestion occurred. Even brief overload conditions could cause sessions to drop. The report explains that vendors began addressing this weakness by making routers intervene in bridged LLC2 exchanges while remaining transparent to end stations, effectively applying routing intelligence to traffic that had previously flowed only through bridges. Tolly benchmarked two early products in this category: the 3Com NETBuilder II and the Cisco Systems 7010 Multiprotocol Router. 

The evaluation focused on two core questions: how much single-session throughput these early DLSw-like routers could sustain across a dedicated T1 link, and whether they could keep 3270 sessions alive across a fully saturated 56Kbit/s WAN connection. In the T1 throughput tests, Tolly used a live SNA/APPC file transfer across paired routers connecting two 16Mbit/s token-ring LANs. Performance was measured with SNA Request/Response Unit sizes of 256, 512, 1,024, and 2,048 bytes. The results were strong. According to the throughput table on page 2, Cisco achieved 1.20, 1.32, 1.36, and 1.44Mbit/s respectively, while 3Com measured 1.22, 1.36, 1.43, and 1.48Mbit/s. With 2,048-byte frames, both products filled more than 93% of the WAN; even with 256-byte frames, typical of 3270 traffic, both exceeded 75% WAN utilization. The report concludes that these routers approached the effective throughput of top-tier token-ring bridges despite the added complexity of DLSw processing. 

Tolly also tested congestion response under difficult 56Kbit/s WAN conditions. In one scenario, the link was flooded with IP traffic. In another, APPC transfers were used again to saturate the line. Rather than enable vendor prioritization features, Tolly deliberately allowed congestion to occur so it could observe whether sessions survived. Every product passed this test: no SNA session was lost, even when congestion temporarily caused sessions to hang. Once the overload subsided, sessions resumed normally on both the client and mainframe sides. That result is especially important for mission-critical mainframe access because it demonstrated that these early DLSw approaches could make SNA sessions significantly more resilient over constrained WAN links. 

The test bed, shown in the diagram on page 3, connected two 16Mbit/s token-ring LANs through a WAN emulator operating at either 56Kbit/s or T1 speeds. APPC end stations used Hewlett-Packard Vectra 486/33U EISA-bus PCs with Olicom token-ring adapters, while the 3270 station used an HP Vectra with a 3Com TokenLink III EISA adapter and Attachmate EXTRA! for Windows 4.0. The mainframe was an IBM 9370 running VM/ESA 1.0, VTAM 3.4, and CMS 7. Instrumentation included a Network General Expert Sniffer, a Wandel & Goltermann DA-30, and an HP Series J2300 Protocol Analyzer for WAN capture, frame verification, and utilization monitoring. Overall, the report presents these results as an encouraging early validation that pre-standard DLSw could deliver both strong throughput and meaningful session protection before formal multivendor standardization was complete.